EEE 435 Principles of Operating Systems

1
EEE 435Principles of Operating Systems

• Basic Memory Management and Page Swapping
• (Modern Operating Systems 4.1 and 4.2)

2
Quick Review
3
Outline

• Memory Management Overview
• Basic Memory Management
• Monoprogramming
• Multiprogramming with fixed partitions
• Swapping
• Management Bitmaps vs. Linked Lists

4
Memory Management Overview

• Why do we need to manage memory?
• How much memory is addressable for a program in
  Windoh!s XP?
• 4 GB (232 bits)
• Do most home computers have that much RAM?
• What do operating systems do when they are
  running programs with memory requirements that
  exceed their capacity?
• Part or all of the programs are swapped to disk

5
Memory Management Overview

• Something is needed to manage the movement of
  programs between disk and memory (or, in general
  to manage the memory hierarchy). The part of the
  operating system that does this is known as the
  ________________

6
Memory Management Overview

• Is multiprogramming worth the hassle of memory
  management?
• Consider a number of processes which wait for the
  CPU an average of 80 of the time
• The probability that n processes will all be
  waiting for I/O at the same time in pn. This
  means that the CPU utilization for n processes is
  1-pn
• Five processes in memory with 80 I/O wait gives
  a CPU utilization of 1-(0.8)5 67 (assuming
  task switch overhead is minimal)
• Much better than the 20 for a single process!

7
Memory Management Overview

• Processes with I/O wait in excess of 80 are not
  uncommon!

8
Basic Memory Management

• First consider two simple systems
• Monoprogramming
• Fixed partitions
• Largely relevant to batch systems
• Once processes are loaded into memory they are
  run to completion

9
Basic Memory Management

• Monoprogramming without swapping or paging
• Simplest scheme possible
• Only one program will be executed at a time
• The OS copies the program from disk to memory and
  executes it. Once finished, the OS is ready to
  accept a new command from the user
• New commands overwrite the old program in memory
  with the new
• Three configurations, see next slide

10
Basic Memory Management

• a) is rarely used anymore
• b) is in use by MP3 players, Palm devices
• c) was the model of early PCs, eg DOS

11
Basic Memory Management

• Multiprogramming with fixed partitions
• To get the benefits of multiprogramming we need
  to be able to have more than one program in
  memory at a time
• Simple solution (for batch systems) divide
  memory into n (possibly unequal) partitions, and
  put arriving jobs into the smallest partition
  that will hold them (or in line for that
  partition)
• Disadvantage since partitions are fixed, any
  space not used by a process is lost

12
Basic Memory Management

• Image shows another disadvantage
• Since we use multiple queues, we have lineups for
  some partitions, and partitions going completely
  unused!
• A single queue usually provides superior service
  (like in a bank)

13
Basic Memory Management

• Many strategies for choosing jobs
• Allow job closest to the front that can fit into
  the free partition to run
• Search the whole input queue for the largest job
  that fits
• Must keep one small partition so small jobs get
  to run!

14
Swapping

• The previous systems were simpler because once
  programs were loaded into memory they were left
  to run to completion
• In a timesharing system we cant choose how many
  processes we keep in memory to keep the CPU
  busy...now the user(s) are dictating how many
  processes are executing
• When not enough main memory exists to hold all
  the active processes we must swap them between
  the disk and main memory...

15
Swapping

• Swapping a process consists of bringing the
  process into memory, from disk, in its entirety.
  The process is executed for some time and then
  placed back on the disk

16
Swapping

• The difference between this system and fixed
  partitions is that the number, location, and size
  of the partitions vary dynamically
• Advantages
• This is a much more flexible solution
• Memory can be better utilized
• Disadvantages
• More complicated to implement
• holes can be left in memory which may need
  compacting to correct

17
Swapping How much memory

• How much memory should be assigned to a process
  when its swapped into memory?
• If a fixed data size can be determined, then
  exactly that much is used
• However, if the process has a heap and/or stack,
  then room for growth will be required to prevent
  having to continually moving the program around
  in memory

18
Swapping How much memory
Data and stack segment

• Data segment

19
Swapping - Management

• How do we keep track of where processes are
  loaded and where space is available to load them?
• Two methods Bitmaps and Linked Lists

20
Swapping - Management

• Memory management with Bitmaps
• Divide memory up into allocation units such as
  4 bytes or up to many kilobytes
• Use a bitmap with 1s to designate an allocated
  area of memory and 0s to represent free space
• How large do we make the allocation unit?

21
Swapping - Management

• Memory management with Bitmaps
• The smaller the allocation unit, the larger the
  corresponding bitmap
• However, even with a 4 byte allocation unit (32
  bits) we only waste 1/33 of the memory
• Larger units stand to waste the end of the last
  unit, eg for a 64KB unit, if we have a program
  that is 65KB, then we waste 63 kilobytes!
• However, we have dropped the bitmap overhead to a
  mere 1/512001 1.9x10-4 of memory

22
Swapping - Management

• Memory management with Bitmaps
• Advantages
• Easy to implement
• Bitmap is a fixed size no matter how many
  programs are in memory
• Disadvantage
• can take a long time to search through the bitmap
  to find a consecutive series of 0s in which to
  place a program

23
Swapping - Management

• Memory management with linked lists
• Use a linked list to keep track of segments
• Segments are either processes or a hole between
  two processes convenient to sort by addresses
• Advantage much less to search through

24
Swapping - Management

• How do we use the linked list to manage processes
  leaving memory?
• Simply combine the new hole with all adjacent
  holes

25
Swapping - Management

• How are new processes placed into memory?
• __________ Find the first empty segment that is
  large enough to hold the process and break it
  into a process segment and a new hole (smaller
  than the previous one)
• __________ Simple improvement on first fit.
  Denote the place where the last program was
  inserted and start looking from there
• Gives slightly worse performance than first fit
  in studies

26
Swapping - Management

• How are new processes placed into memory?
• __________ Instead of taking a near fit and
  breaking it into a process segment and a very
  tiny, unusable hole, find the largest available
  hole to leave maximum space for other processes
• Simulation has shown this to be a truly awful
  choice
• __________ Keep common entries such as 4KB, 8KB,
  etc in a separate table for easy hole location
• Good for allocation, but slower for de-allocation
  as multiple lists must be reconciled

27
Swapping - Management

• Final note on lists you could keep separate
  lists for processes and holes
• Speeds up looking for a hole!
• Allows you to sort holes by size for even quicker
  allocation!!
• Lets you use the actual memory holes themselves
  to hold the pointer information from one hole to
  the next!!!
• More complicated on de-allocation as your memory
  has to be placed in the correct spot on the other
  list...

28
Quiz Time!

• Questions?